AI Is Nothing New: Moore’s Law Finally Gave Artificial Intelligence the Processing Power It Needed

Artificial intelligence can feel as though it suddenly appeared.

ChatGPT arrived. Generative AI exploded. Organizations began discussing AI governance, AI auditing, AI fraud detection, automated data analysis, machine learning, and large language models almost overnight.

But AI itself is not new.

The basic intellectual foundations of artificial intelligence are roughly 70 years old. The famous Dartmouth Summer Research Project on Artificial Intelligence was held in 1956, and John McCarthy is credited with coining the term “artificial intelligence” in connection with that work.

What changed was not the fundamental desire to make computers reason, recognize patterns, learn, and solve problems.

What changed was our ability to provide computers with the processing power, memory, data, storage, and scale necessary to make increasingly sophisticated AI practical.

And one of the best ways to understand that transformation is through the remarkable prediction made by Gordon Moore.

One historical correction is important: Moore did not make his famous prediction in the 1950s. He published it in 1965. His observation grew from semiconductor advances that began in the late 1950s.

AI Was Born in the 1950s

The history matters.

In 1955, John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon proposed a summer research project at Dartmouth based on the idea that aspects of human intelligence could potentially be described precisely enough for a machine to simulate them.

The workshop took place during the summer of 1956 and helped establish artificial intelligence as an academic field.

Think about that.

In the 1950s, researchers were already considering whether computers could:

• Learn

• Reason

• Solve problems

• Use language

• Recognize patterns

• Simulate aspects of intelligence

Those ideas are remarkably familiar to anyone discussing AI today.

The problem was that the computers available to those researchers were extraordinarily limited compared with modern systems.

They had the ideas.

They did not yet have the computational infrastructure.

Then Came Gordon Moore

Gordon Moore was working at Fairchild Semiconductor when he noticed something extraordinary happening in integrated circuits.

In his 1965 article, “Cramming More Components onto Integrated Circuits,” Moore observed the rapid increase in the number of components that could economically be placed on an integrated circuit.

He projected the trend forward.

Moore initially predicted roughly an annual doubling in component density over the following decade. In 1975, he revised the future rate to approximately doubling every two years.

That observation became known as:

Moore’s Law

It isn't a physical law like gravity.

It was an observation and extrapolation concerning semiconductor development.

But it became extraordinarily influential.

Intel describes the modern shorthand this way: the number of transistors on a chip would roughly double every two years with minimal increase in cost.

The implications were enormous:

More transistors

↓

More computational capability

↓

Smaller systems

↓

Lower relative cost

↓

More applications

↓

More data

↓

More sophisticated software

↓

More sophisticated AI

That technological progression helps explain why ideas conceived in the 1950s could become commercially transformative decades later.

Moore Wasn't Predicting ChatGPT

This distinction is important.

Gordon Moore did not predict artificial intelligence in the form we know today.

He was describing semiconductor economics and integrated-circuit complexity.

But his observation pointed toward the technological foundation that would make modern computing possible.

Moore himself was trying to demonstrate that integrated circuits would eventually make electronics dramatically cheaper. His 1965 article also anticipated that integrated circuits would lead to more powerful computers and many new applications.

That is the connection to AI.

Artificial intelligence requires computation.

More sophisticated artificial intelligence requires enormous amounts of computation.

The exponential improvement in semiconductor technology made previously impractical computations increasingly possible.

Think About 60 Years of Doubling

This is where Moore's observation becomes difficult to comprehend intuitively.

Human beings naturally think linearly.

If something improves from 1 to 2, we tend to imagine:

1

2

3

4

5

But exponential improvement looks more like:

1

2

4

8

16

32

64

128

256

512

1,024

and so forth.

Keep doubling and the numbers become enormous.

The Computer History Museum notes that Moore originally extrapolated from the relatively tiny integrated circuits of the early 1960s to approximately 65,000 components per chip by 1975. That prediction proved remarkably close to what semiconductor technology ultimately achieved.

By the 2010s, chips contained billions of transistors.

That is an extraordinary transformation in computing capability.

AI Was Waiting for the Hardware to Catch Up

This is a useful way for auditors and financial professionals to understand the history of AI.

The concept did not suddenly emerge in 2022.

Instead, several technological curves eventually converged.

We developed dramatically more powerful processors.

Memory became vastly larger and cheaper.

Storage became inexpensive.

Networks became faster.

Cloud computing made massive computational resources accessible.

Organizations accumulated enormous digital datasets.

Graphics processors and specialized AI accelerators enabled highly parallel computations.

Algorithms improved.

Then all of these capabilities began working together.

The result was not simply more computing.

It crossed thresholds that allowed computers to perform tasks that previously seemed impractical.

The Difference Between an Old Idea and a Practical Technology

Consider voice recognition.

The idea of getting a computer to recognize human speech is old.

But recognizing speech effectively requires substantial computational resources and data.

The same applies to:

• Image recognition

• Language translation

• Fraud detection

• Predictive analytics

• Natural-language processing

• Document analysis

• Generative AI

The underlying objective may have existed for decades.

The technology needed to perform the task economically and at scale did not.

That distinction is important:

An invention can exist conceptually long before the supporting technology makes it commercially useful.

AI provides a remarkable example.

ChatGPT Is the Visible Result of a Much Longer Technological Evolution

When many professionals first used ChatGPT, the experience seemed revolutionary.

You could ask a computer:

“Explain this accounting standard.”

Then:

“Compare it with this other standard.”

Then:

“Identify the risks.”

Then:

“Develop an audit program.”

Then:

“Analyze this spreadsheet.”

Then:

“Draft the audit finding.”

Then:

“Rewrite the finding for the Audit Committee.”

To someone who spent a career using traditional computer systems, this feels completely different.

And at the user-interface level, it is.

But underneath that experience sits decades of accumulated technological development.

I would describe it this way:

1950s — The AI Vision

↓

1960s — Integrated Circuits and Moore's Observation

↓

1970s–1980s — Increasing Computing Power

↓

1990s — Personal Computing and the Internet

↓

2000s — Massive Digital Data and Cloud Infrastructure

↓

2010s — GPUs, Deep Learning and Large-Scale Machine Learning

↓

2020s — Generative AI and Large Language Models

The revolution didn't start with ChatGPT.

ChatGPT made the revolution visible to millions of ordinary users.

Why Did AI Seem to Arrive So Suddenly?

Because exponential technological development often looks unimpressive until it reaches a threshold.

Imagine a system that is only 60% accurate.

Interesting.

At 70%, perhaps useful for experimentation.

At 80%, it starts becoming valuable.

At 90%, businesses begin adopting it.

At 95%, workflows can begin changing.

The underlying technology may have been improving for decades.

But users notice it when it crosses the threshold from:

Interesting

to:

Useful.

Generative AI crossed that threshold for millions of knowledge workers.

Auditors Have Seen This Before

There is an interesting parallel in auditing.

Auditors have talked about continuous auditing for decades.

Why wasn't every audit continuous 30 years ago?

Partly because the technological infrastructure wasn't ready.

The auditor needed:

• Digital transactions

• Electronic records

• Accessible databases

• Processing power

• Analytical software

• Storage

• System integration

• Automated controls

As those capabilities improved, continuous auditing became increasingly practical.

AI follows the same pattern.

The idea may be old.

The economics changed.

Processing Power Changed What the Auditor Can Do

Consider a traditional audit.

The auditor has a population of: 2,500,000 transactions.

Historically, the auditor might select: 60 transactions.

Why?

Not because the auditor believes the other 2,499,940 transactions are irrelevant.

Sampling was partly a practical response to limited resources.

Now consider an AI- and analytics-enabled audit environment.

Technology can potentially analyze the entire population and identify:

• Duplicates

• Unusual amounts

• Transactions below approval thresholds

• Weekend transactions

• Unusual vendors

• Employee/vendor address matches

• Unexpected journal entries

• Round-dollar transactions

• Changes to bank accounts

• Unusual approval patterns

Now the auditor can ask a different question.

Instead of:

“Which 60 transactions should I sample?”

the auditor can ask:

“Which transactions in the entire population exhibit characteristics that deserve investigation?”

That is a profound change in audit methodology.

AI Is the Latest Tool in a Long History of Audit Technology

Auditors have continually adopted new tools.

Paper ledgers

became

Mainframe accounting systems

which became

Personal computers

which produced

Spreadsheets

followed by

Audit analytics

then

Continuous monitoring

and now

Artificial intelligence.

Every technological transition changed the auditor's tradecraft.

AI should be viewed the same way.

It is a tool.

A very powerful tool—but still a tool.

The Auditor's Brain Hasn't Been Replaced

This is where some of the AI hype goes too far.

Computing power does not eliminate professional judgment.

AI can tell the auditor:

“These 147 transactions are unusual.”

The auditor still needs to ask:

Why?

AI can identify a correlation.

The auditor determines whether it represents risk.

AI can draft a finding.

The auditor determines whether the evidence supports it.

AI can recommend controls.

The auditor evaluates whether those controls are appropriate for the organization.

AI can summarize an interview.

The auditor evaluates credibility, behavior, contradictions, and context.

AI changes the processing capability available to the auditor.

It does not transfer the auditor's professional responsibility to the machine.

Moore's Law Is Also a Warning to Auditors

There is another side to this story.

If computing capability continues improving rapidly, auditors cannot assume today's AI capabilities define tomorrow's AI.

Think about how quickly earlier technologies changed.

A computer purchased five years later wasn't merely a slightly improved version of the old computer.

Eventually, it could do things the earlier machine simply could not do economically.

That means auditors need to think differently about AI risk.

Don't ask only:

“What can AI do today?”

Ask:

“What happens to our organization when AI capabilities become dramatically better and cheaper?”

That creates new risks involving:

• Fraud

• Cybersecurity

• Deepfakes

• Identity

• Financial reporting

• Automated decision-making

• Data privacy

• Model risk

• Intellectual property

• Third-party systems

• Internal controls

The same computational progress that empowers auditors also empowers fraudsters.

Moore's Law Meets the Fraudster

Suppose AI allows an auditor to analyze millions of transactions.

Excellent.

The fraudster has access to AI too.

Suppose AI allows an auditor to create sophisticated analytical models.

The criminal can use AI to develop sophisticated attacks.

Suppose AI helps an organization communicate with customers.

AI can also help criminals impersonate those customers.

Technology creates capability.

Capability itself is neither a control nor a risk.

How people use the capability determines the consequence.

That is why auditors need to understand the technology rather than simply admire it.

The Real AI Revolution Is Economics

One of Moore's most important insights wasn't simply that chips would contain more components.

It was that integrated electronics would become increasingly economical.

That is enormously relevant to AI.

A technology may technically be possible for decades and still have little business impact if it costs $10 million to perform a task.

Make that same capability available for $100,000 and the market changes.

Make it available for $1,000 and small organizations adopt it.

Make it available for $20 per month and virtually every knowledge worker can experiment with it.

That is when technology changes business.

AI's impact is therefore not simply about intelligence.

It is about the rapidly declining cost of computational capability.

What Does This Mean for Internal Audit?

Internal Audit should stop treating AI as an exotic technology project.

AI belongs in the audit universe.

Auditors should consider three separate questions.

First, how can Internal Audit use AI?

Risk assessment, research, data analysis, audit planning, interview preparation, workpaper development, root-cause analysis, report writing and monitoring.

Second, how is the organization using AI?

Management processes, financial reporting, HR, procurement, customer service, cybersecurity, compliance and decision-making.

Third, what new risks does AI create?

Data integrity, unauthorized use, hallucinations, privacy, bias, model governance, cybersecurity, third-party risk, intellectual property and overreliance.

Those three perspectives give Internal Audit a much stronger framework than simply asking:

“Do we have an AI policy?”

Don't Be Intimidated by AI

There is perhaps a more important lesson for experienced auditors.

Don't assume AI belongs exclusively to young programmers, data scientists, or technology auditors.

If you understand:

• Objectives

• Risks

• Controls

• Evidence

• Fraud

• Business processes

• Professional skepticism

you already possess much of the knowledge needed to use AI effectively in auditing.

The machine provides computational capability.

The auditor provides the question.

And the quality of the question remains extraordinarily important.

AI Doesn't Change the Fundamental Audit Model

Whether an auditor is using a yellow pad, Excel, audit analytics, or artificial intelligence, the fundamental logic remains:

OBJECTIVE

↓

RISK

↓

CONTROL

↓

EVIDENCE

↓

ANALYSIS

↓

CONCLUSION

↓

CORRECTIVE ACTION

AI can dramatically accelerate the middle of that process.

It can analyze more evidence.

Find more relationships.

Generate more alternatives.

Search more information.

Draft more quickly.

But the auditor remains responsible for determining whether the conclusion is reasonable.

Gordon Moore Helps Explain Why This Is Happening Now

So when someone asks:

“Where did all this AI suddenly come from?”

My answer is: It didn't suddenly come from anywhere.

The field of artificial intelligence dates to the 1950s.

Gordon Moore then observed in 1965 that semiconductor complexity was increasing exponentially, an observation that eventually became the semiconductor industry's famous Moore's Law.

For decades, processors became vastly more capable, computing became cheaper, storage expanded, networks accelerated, digital information exploded, algorithms improved, and massive computational infrastructure became available.

Eventually those trends reached a point where ideas researchers had discussed for decades became practical tools that ordinary professionals could use.

AI isn't new.

Our ability to deploy extraordinarily powerful AI economically and at enormous scale is new.

And that distinction matters.

The Bottom Line: AI Is Evolution at Exponential Speed

The history can be summarized simply:

The idea came first.

The algorithms developed.

The data accumulated.

The chips became powerful enough.

The economics became practical.

Then the applications exploded.

That is why auditors should not view AI as magic.

It is another stage in the extraordinary evolution of computing.

Gordon Moore did not predict ChatGPT.

But his 1965 observation helps explain the technological trajectory that ultimately made systems like it possible.

And perhaps the most important lesson from Moore's Law for today's auditor is not about the past.

It is about the future.

If computational capability has changed this much over the last several decades, we should expect the tools available to auditors—and to the people we audit—to continue changing dramatically.

The question for Internal Audit is therefore not:

“Will AI affect auditing?”

It already has.

The better question is:

“How quickly can auditors learn to use this new computational power while maintaining the professional judgment, skepticism, ethics, and evidence standards that technology cannot replace?”

That is the real AI challenge for the auditing profession.